Macrophages serve several roles in tissue maintenance and inflammatory response in both normal and diseased tissues (Gordon and Plüddemann, 2017). There is limited understanding of functional differences in macrophages across these tissue sites and of macrophage involvement in tissue dysfunction. Additionally, macrophage plasticity results in multiple unique functional phenotypes, presenting further challenges for assessing macrophage function (Shapouri-Moghaddam et al., 2018). Standard functional assays (e.g., flow cytometry, PCR, ELISA, and histology) require destructive sample preparation, which limit assessment of macrophage behavior in vivo (Chattopadhyay et al., 2014). Non-destructive methods of monitoring macrophage function are needed to better understand macrophage plasticity and the role of macrophages in regulating tissue homeostasis and disease pathogenesis.

Tissue niche conditions, such as cytokine/chemokine secretion, nutrient availability, and collagen organization, drive macrophages to adopt various functional phenotypes (Epelman et al., 2014; Lavin et al., 2014; Sridharan et al., 2019; Sapudom et al., 2020).Generally, tissue-resident macrophages serve to protect against invading pathogens and promote tissue repair, independent of tissue origin (Gordon and Plüddemann, 2017). Previous studies have shown that functional gene expression differs between macrophages originating from distinct tissue sites, and these macrophages can be functionally reprogrammed upon trafficking to other tissues (Lavin et al., 2014). Additionally, tumor-infiltrating macrophages are specifically directed to promote tumor progression by signaling tumor cell proliferation, motility, and angiogenesis while suppressing further immune recruitment (De Palma and Lewis, 2013). However, comparisons between distinct macrophage populations across normal and diseased tissue sites are limited due to constraints of functional assays (De Palma and Lewis, 2013; Panni et al., 2013).

Changes in macrophage phenotype and function have been directly correlated to their metabolic demands (Stout et al., 2005; De Palma and Lewis, 2013; Panni et al., 2013). Macrophages undergo metabolic switching between oxidative and glycolytic metabolism to fuel anti- or pro-inflammatory processes, respectively (Diskin and Pålsson-McDermott, 2018). Energy demands of tissue-resident macrophages are regulated by the tissue microenvironment, though characterization of macrophage metabolic profiles across specific tissues are limited (Caputa et al., 2019). Previous studies suggest that tissue-resident macrophages require stable upregulation of both glycolysis and oxidative phosphorylation to promote survival and proliferation in tissue (Van Den Bossche et al., 2015). Tumor-infiltrating macrophages are primarily characterized by an anti-inflammatory phenotype relying on fatty acid oxidation and oxidative phosphorylation (Panni et al., 2013). Conversely, highly glycolytic macrophage populations have also been observed invading tumors, demonstrating metabolic heterogeneity associated with tumor-infiltrating macrophages (Vitale et al., 2019). Tools for monitoring macrophage metabolic changes in vivo may better visualize this heterogeneity and inform on cell-level behavior of macrophage populations and interactions with the tissue microenvironment.

Two-photon microscopy of reduced nicotinamide adenine dinucleotide (phosphate) (NAD(P)H) and oxidized flavin adenine dinucleotide (FAD), or “metabolic autofluorescence imaging,” monitors metabolic activity on a single-cell level. NAD(P)H and FAD are an electron donor and acceptor, respectively, involved in metabolic reactions across numerous pathways (Chance et al., 1979; Walsh et al., 2013; Blacker and Duchen, 2016). NAD(P)H and FAD fluorescence intensities report on their respective abundance within cells (Chance et al., 1979; Blacker and Duchen, 2016). Furthermore, the optical redox ratio, defined as the ratio of fluorescence intensity of NAD(P)H to FAD, provides a quantitative measurement of the relative oxidation-reduction state of individual cells (Lakowicz et al., 1992; Georgakoudi and Quinn, 2012; Walsh et al., 2013; Blacker and Duchen, 2016). The fluorescence lifetimes of free and protein-bound NAD(P)H and FAD are distinct and can be recovered with a bi-exponential decay model (Lakowicz et al., 1992; Georgakoudi and Quinn, 2012; Quinn et al., 2013; Walsh et al., 2013). Therefore, fluorescence lifetime imaging microscopy (FLIM) of NAD(P)H and FAD reflects the protein-binding activities of these metabolic co-enzymes in the cell. Previous studies have demonstrated that in vivo tumor-associated macrophages have high FAD intensities that can discriminate these macrophages from other cell types (Szulczewski et al., 2016; Li and Liu, 2018). However, tissue-specific macrophage metabolism and the metabolic heterogeneity of macrophages within a tissue have not been thoroughly characterized in vivo due to a lack of appropriate tools. NAD(P)H and FAD autofluorescence have been previously used to distinguish metabolically diverse cell sub-populations (Kilarski et al., 2013; Ghesquière et al., 2014; Tong et al., 2015; Walsh and Skala, 2015; Alfonso-García et al., 2016; Shah et al., 2017; Heaster et al., 2018, 2020; You et al., 2018; Jones et al., 2020; Li et al., 2020; Smokelin et al., 2020), which supports the use of metabolic autofluorescence imaging to investigate macrophage heterogeneity in vivo. This study demonstrates that metabolic autofluorescence imaging can quantify metabolic heterogeneity between macrophages within normal and cancerous mouse tissues in vivo.

Representative images of optical redox ratio, NAD(P)H τ_m, and FAD τ_m depict qualitative differences in the metabolic activity and spatial organization of dermal macrophages and tumor-infiltrating macrophages (Figure 2B). Macrophages (mCherry+ cells segmented with geodesic reconstruction) localize around hair follicles in the ear dermis, consistent with previous observations (Palero et al., 2007). Conversely, tumor-associated macrophages are randomly distributed across the tissue (Figure 2B). These results indicate that metabolic autofluorescence imaging can visualize metabolic differences in macrophages in vivo with single-cell resolution across multiple tissue types.

Using the composite imaging and single cell segmentation shown in Figure 2, quantitative analyses of the full data set for all mice enables comparisons of tumor and dermal macrophage autofluorescence in vivo (Figure 3). Tumor-infiltrating macrophages exhibited a small but significant decrease in optical redox ratio compared to macrophages within dermal tissue (Figure 3A), consistent with a metabolic preference for oxidative phosphorylation (Qian and Pollard, 2010; Panni et al., 2013). Furthermore, dermal macrophages displayed higher NAD(P)H τ_m and FAD τ_m compared to tumor-infiltrating macrophages (Figures 3B,C). The distributions of redox and lifetime values represented by the violin plots reveal notable metabolic heterogeneity in macrophages within both tissue types (Figures 3A–C). Dermal macrophages demonstrate greater metabolic heterogeneity compared to tumor-associated macrophages, as shown in the violin plots and coefficients of variation for all autofluorescence variables (Figure 3D). Macrophage variability is highest for the optical redox ratio, regardless of tissue type (CV: 30.37% – tumor; 108.05% – ear dermis). Ultimately, these quantitative trends illustrate the sensitivity of metabolic autofluorescence imaging to heterogeneous macrophage populations in vivo.

Immunofluorescence was used to confirm the heterogeneity observed with in vivo autofluorescence imaging (Figure 4). Collectively, dermal tissue yielded much fewer total macrophages (mCherry+ cells) than tumor tissue, supporting the fewer number of macrophages imaged in the dermis compared to the tumor with in vivo autofluorescence imaging (Figure 4A). Macrophages positive for CD206, commonly associated with M2-like macrophages that prefer oxidative metabolism were abundant in tumors (Figure 4A). Significant CD206 staining was also observed for dermal macrophages but yielded much sparser expression than tumor-associated macrophages (Figure 4A). These CD206+mCherry+ macrophages (yellow, Figure 4B inlay) were also more prevalent surrounding hair follicles (white arrows), similar to the localization observed in metabolic autofluorescence images. Despite notable CD86 expression in tumor-associated macrophages, limited numbers of CD86+mCherry+ macrophages (purple, Figure 4B inlay) were detected across either tissue type. These qualitative observations are apparent in the quantification of relative proportions of CD86+ mCherry+ or CD206+ mCherry+ cells from immunofluorescence images (Figure 4C and Supplementary Figure 2). Tumor-infiltrating macrophages were observed to be primarily CD206+ (52.2% of mCherry+ cells vs. 8.5% CD86+ mCherry+ cells, p < 0.0001; Figure 4C), while dermal macrophages exhibited a more heterogenous mixture of CD206+ and CD86+ phenotypes (35.9% of CD206+ mCherry+ cells vs. 20.6% CD86+ mCherry+ cells, p > 0.05; Figure 4C). Note that some mCherry+ cells do not stain for CD86 or CD206 [e.g., non-M1/M2-type macrophages (Aras and Zaidi, 2017)], so the total percentage in Figure 4C does not reach 100%. Overall, immunofluorescence staining of excised mouse ear dermis and tumors confirmed that heterogeneous macrophage phenotypes contribute to the heterogenous macrophage autofluorescence in both dermal and tumor tissues. These immunofluorescence data also support greater heterogeneity in dermal macrophage phenotype compared to tumor macrophage phenotype, consistent with autofluorescence imaging.

This study establishes metabolic autofluorescence imaging as an effective approach for non-destructive, in vivo assessment of macrophage metabolism across normal and cancerous tissue in immunocompetent mouse models. Distinct tissue microenvironments have been previously shown to alter macrophage phenotype and function, providing innate perturbation of macrophage metabolism in vivo (Stout et al., 2005). Mouse ear dermis served as an optimal site to measure non-diseased tissue for comparison with engrafted tumors in a single animal, without the need for multiple surgeries. Autofluorescence imaging enabled quantitative assessment of metabolic heterogeneity within infiltrating macrophage populations across these tissues. Standard immunofluorescence confirmed the observations from metabolic autofluorescence imaging of macrophages within both dermal and tumor tissues. Collectively, these findings support the use of in vivo metabolic autofluorescence imaging to study relationships between macrophage behavior and the local tissue microenvironment.

The combination of intravital autofluorescence imaging and reporter mouse models ensures that the cells of interest (i.e., macrophages) can be accurately identified in vivo. Here, we used a previously established macrophage reporter model generated from crossbreeding Lyz2-Cre mice and mice harboring a floxed mCherry under the Csfr1 promoter (Schreiber et al., 2013; Hoffmann and Ponik, 2020). These mice express mCherry restricted to blood-circulating monocytes and those that have infiltrated into tissue becoming macrophages (Das et al., 2015). Previous studies injected fluorescently labeled antibodies into tissues to identify macrophages and then co-registered autofluorescence signals within these antibody-labeled macrophages in vivo (Szulczewski et al., 2016). This approach was limited by issues with in vivo antibody delivery, non-specific antibody binding, and spectral overlap between the fluorescent label and FAD autofluorescence. The use of fluorescent reporter mice in the current study provided macrophage-specific labels while circumventing these issues of antibody delivery and non-specific binding. The mCherry reporter used in the current study also did not interfere with NAD(P)H and FAD autofluorescence, similarly observed in previous in vivo autofluorescence studies (Hoffmann and Ponik, 2020; Miskolci et al., 2020). Identifying macrophage-specific autofluorescence features without fluorescent reporters remains challenging to due to substantial macrophage heterogeneity in vivo. However, this model could serve as a basis for cultivating reference in vivo datasets to train machine learning classifiers for unlabeled macrophages (Rostam et al., 2017; Christiansen et al., 2018; Pavillon et al., 2018). Overall, the combination of macrophage-specific mCherry reporter mice and two-photon autofluorescence imaging provide a convenient method to monitor macrophage behavior in vivo.

Macrophages traffic to specific tissues and undergo phenotypic switching in response to tissue-secreted biochemical signals (Aloysius et al., 2020). Dermal macrophages comprise a large proportion of cells (∼60%) in the mouse ear dermis, primarily of glycolytic, M1-like phenotype (Dupasquier et al., 2004; Italiani and Boraschi, 2014). The increased optical redox ratio of dermal macrophages reflects this abundant functional phenotype in the mouse dermis (Figure 3A). The increased optical redox ratio and FAD τ_m observed for dermal macrophages (Figures 3A,C) are also consistent with increased optical redox ratio and FAD τ_m in previous studies of 2D mouse macrophage cell lines stimulated to glycolytic, M1-like phenotypes (compared to naïve and M2-like macrophages) (Heaster et al., 2020). Similarly, macrophages are highly abundant in mouse pancreatic tumors (Sindrilaru et al., 2011; Ino et al., 2013). However, tumors commonly promote immunosuppressive function and oxidative metabolism in infiltrating macrophages to promote tumor survival, associated with M2 phenotype and function (Qian and Pollard, 2010). This reported preference for oxidative metabolism in tumor-infiltrating macrophages agrees with our measured decrease in optical redox ratio and FAD τ_m in tumor-infiltrating macrophages compared to dermal macrophages (Figures 3A,C). Our autofluorescence measurements in pancreatic tumor-infiltrating macrophages (Figures 2, 3) are also comparable to previous studies that show low NAD(P)H τ_m and high FAD intensity in tumor-infiltrating macrophages in vivo (Szulczewski et al., 2016). Small differences in absolute lifetime values reported here and in the previous in vivo autofluorescence study (Szulczewski et al., 2016) may be attributed to differences in tumor site (breast vs. pancreatic cancer) and tumor model (spontaneous vs. engrafted), which are known to alter cell autofluorescence (Heaster et al., 2020). Overall, these results demonstrate that metabolic autofluorescence imaging can detect heterogenous macrophage populations across normal and cancerous tissue based on unique macrophage metabolic activities.

Finally, phenotypic and functional heterogeneity captured from metabolic autofluorescence was confirmed with immunofluorescence staining of known macrophage markers. Diversity in metabolic autofluorescence was observed for both tumor-infiltrating macrophages and dermal macrophages (Figure 3D). Immunofluorescence confirmed this diversity, as mixed CD206 and CD86 expression was observed for both tumor-infiltrating macrophages and dermal macrophages (Figure 4). This heterogeneity is consistent with previous studies of tumor-associated macrophage diversity measured with flow cytometry in this tumor model (Biswas and Mantovani, 2012; Urs, 2019). Similarly, previous studies have measured diverse macrophage populations from immunofluorescence characterization of mouse skin comparable to the CD86/CD206 staining observed for ear dermis tissue (Sindrilaru et al., 2011; Lee and Prisby, 2017). This agreement with immunofluorescence supports autofluorescence imaging of NAD(P)H and FAD as a label-free in vivo method to monitor spatial and temporal variations in both macrophage metabolism and phenotype.

Here, we have shown that intravital metabolic autofluorescence imaging can resolve dynamic macrophage function within native tissue microenvironments, confirmed with fluorescent reporter mouse models. This provides an attractive method to assess macrophage behavior throughout disease progression and in response to treatment, which could aid in new drug development to target macrophages in cancer, wound healing, and a variety of other diseases. While autofluorescence FLIM alone is not sufficient to specify a biological mechanism, the in vivo macrophage imaging techniques described in this paper can be used to guide independent measurements of metabolism. The methods described here are also amenable to time-course measurements of macrophage movement and autofluorescence over hours and days within the same animal in vivo. Longer-term studies require alternative procedures (e.g., specialized vital regulation and imaging windows) though could provide further insight into the evolution of macrophage function in vivo (Ingman et al., 2006; Szulczewski et al., 2016; Entenberg et al., 2017; Dawson et al., 2021). These methods could also identify novel approaches to alter tissue state to influence macrophage function. Ultimately, this imaging approach could elucidate the dynamic role of tissue-specific macrophage populations in vivo.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Institutional Animal Care and Use Committee, the University of Wisconsin–Madison.

TH designed the experiments, collected the data, designed and performed the analysis, interpreted the data, generated the figures, and drafted the manuscript for this study. AH assisted in performing the analysis, interpreting the data, and editing the manuscript. PS and MC helped to design the experiments and analysis, interpreted the data, and edited the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.